T-test and univariate logistic regression can be used to find the significant feature.
In t-test analysis, we have checked the significance difference between two group of data (P-value < 0.05 show that this feature significantly differentiate the classes); in univariate logistic regression, we have considered the p-value of coefficient (P-value
Hi,
One of the important differences between the ttest and Logestic regression is that the DV in ttest should be a continous measured variale, but in LR the DV is Dichotomous.
I totally agree with Prof Muayyad Ahmad and Ertuğrul Şahin. But sir, can we use both for feature selection?
1st we use t-test analysis to select those feature that significantly differentiate the classes, then we use the univariate logistic regression to check whether the model developed using selected set of significant feature is significant or not. Based on these two analysis we select and reject the hypothesis.
The t-test or the univariate logistic regressions can be used for an initial screening or filtering. They will give pretty similar answers.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Energy